A comparative assessment of artificial neural network and regression models to predict mechanical properties of continuously cooled low carbon steels: an external data analysis approach

Yıl 2024, Cilt: 4 Sayı: 2, 495 - 513, 31.07.2024

Emre Alan , İsmail İrfan Ayhan , Bilgehan Ögel Deniz Uzunsoy

https://doi.org/10.61112/jiens.1445518

Öz

In this study, mechanical properties of continuously cooled low carbon steels were predicted via Artificial Neural Network (ANN) and Multiple Linear Regression (MLR) models. Unlike the previous studies, laboratory scaled self-generated data that consists of chemical compositions and cooling rates were used as input while yield strength (YS), ultimate tensile strength (UTS) and total elongation (TE) were served as target data. The prediction performances of the models were compared by applying new data set extracted from external sources like previously studied research papers, thesis or dissertations. A better agreement between predicted and actual data was achieved with ANN model. Additionally, the response of ANN model to new external data resulted in lower prediction errors even the data has one or more input value that is not included in the range of training data set. Unlike ANN model, MLR model shows a significant decrease in prediction accuracy when input data has non-uniform distribution or target data takes place in relatively narrow range. In general, it was shown that ANN model trained with self-generated data can be used as an efficient tool to estimate mechanical properties of continuously cooled low carbon steels that are produced with various conditions, even for the phenomena between input and output is complex and data distribution is non-uniform.

Anahtar Kelimeler

Artificial neural network, Multiple linear regression, Continuous cooling, Low carbon steel, Mechanical properties.

Destekleyen Kurum

ÇEMTAŞ Çelik Mak. San. ve Tic. A. Ş., Bursa Technical University

Proje Numarası

TÜBİTAK 1002-A Project No.: 222M041

Sürekli soğutma ile üretilen düşük karbonlu çeliklerin mekanik özelliklerinin tahmininde yapay sinir ağları ve çoklu regresyon analizi karşılaştırması: dış veri doğrulaması yaklaşımı

Öz

Bu çalışmada, sürekli soğutma ile üretilen düşük karbonlu çeliklerin mekanik özellikleri Yapay Sinir Ağları (ANN) ve Çoklu Regresyon Analizi (MLR) modelleri kullanılarak tahmin edilmiştir. Oluşturulan modellerde çelik kimyasal kompozisyonları ve soğutma hızları (CR) girdi parametreleri olarak kullanılıp akma mukavemetleri (YS), çekme mukavemetleri (UTS) ve uzama değerleri (TEl) hesaplanmıştır. Önceki çalışmalardan farklı olarak, kullanılan parametrelerin tamamı laboratuvar ölçekli üretimlerden elde edilmiş ve modellerin performansları harici olarak farklı kaynaklarda yer alan veriler ile karşılaştırmalı değerlendirilmiştir. Yapılan incelemede ANN modeli ile tahmin edilen değerlerin MLR modeline göre daha düşük hata oranlarında gerçekleştiği gözlemlenmiştir. Ayrıca, dış kaynaklardan alınan veriler ile yapılan tahminlerde ANN modelinin veri setinin düzensiz olduğu durumlarda bile daha yüksek performans sağladığı hesaplanmıştır. MLR modelinde ise kullanılan verilerin düzgün dağılım göstermeyip belirli bir aralıkta kümelendiği veya girdi veri setinin dışında yer aldığı durumlarda performansında önemli bir düşüş oluştuğu belirlenmiştir. Genel olarak değerlendirildiğinde, laboratuvar ölçekli olarak üretilen ve sürekli soğutma uygulanan düşük karbonlu çeliklerin mekanik özelliklerinin tahmini için yapılan karşılaştırmalı incelemede, kullanılan veri setinin düzensiz olduğu durumlarda dahi ANN modeline ait tahminlerin MLR modeline göre daha düşük hata oranlarında elde edildiği belirlenmiştir.

Anahtar Kelimeler

Yapay Sinir Ağları, Çoklu Regresyon Analizi, Sürekli Soğutma, Düşük Karbonlu Çelikler, Çeliklerin Mekanik Özellikleri

Proje Numarası

TÜBİTAK 1002-A Project No.: 222M041

Kaynakça

• Schmitt JH, Iung T (2018) New developments of advanced high-strength steels for automotive applications. Comptes Rendus. Physique 19(8):641-56.
• Lesch C, Kwiaton N, Klose FB (2017) Advanced high strength steels (AHSS) for automotive applications− tailored properties by smart microstructural adjustments. Steel Res. Int. 88(10):1700210.
• Kwon O, Lee KY, Kim GS, Chin KG (2010) New trends in advanced high strength steel developments for automotive application. Mater. Sci. Forum 8(638):136-141).
• Kučerová L, Jirková H, Mašek B (2016) Influence of Nb micro-alloying on TRIP steels treated by continuous cooling process. Manuf. Technol. 16(1):145-9.
• Hasan SM, Ghosh M, Chakrabarti D, Singh SB (2020) Development of continuously cooled low-carbon, low-alloy, high strength carbide-free bainitic rail steels. Mater. Sci. Eng. A 771:138590.
• Gomez G, Pérez T, Bhadeshia HK (2008) Strong bainitic steels by continuous cooling transformation. New Dev. Metall. Appl. High Strength Steels 1:571-82.
• Gigović-Gekić A, Oruč M, Avdušinović H, Sunulahpašić R (2014) Regression analysis of the influence of a chemical composition on the mechanical properties of the steel nitronic 60. Mater Tehnol 48(3):433–437.
• Chang J, Wang Z, Xiao T, Xin X (2018) Statistical Analysis of the Effects of Mn and Cr Contents on Mechanical Properties of Deformed Steel Bar. Proceedings of the 2018 International Conference on Mathematics, Modelling, Simulation and Algorithms (MMSA 2018), Chengdu, China, 25-26 March 2018. pp. 418-423.
• Tumrate CS, Chowdhury SR, Mishra D (2021) Development of Regression Model to Predicting Yield Strength for Different Steel Grades. IOP Conf. Ser.: Earth Environ. Sci 796(1):012033.
• Jones DM, Watton J, Brown KJ (2005) Comparison of hot rolled steel mechanical property prediction models using linear multiple regression, non-linear multiple regression and non-linear artificial neural networks. Ironmaking Steelmaking 32(5):435-42.
• Sankar IB, Rao KM, Gopalakrishna A (2010) Optimization of steel bars subjected to Tempcore process using regression analysis and harmony search algorithm. Pak J Sci Ind Res 69:266-270.
• Quiza R, Figueira L, Paulo Davim J (2008) Comparing statistical models and artificial neural networks on predicting the tool wear in hard machining D2 AISI steel. Int J Adv Manuf Technol 37:641.
• Mukherjee I, Ray PK (2006) A review of optimization techniques in metal cutting processes. Comput Ind Eng 50(1-2):15-34.
• Derin B, Alan E, Suzuki M, Tanaka T (2016) Phosphate, phosphide, nitride and carbide capacity predictions of Molten melts by using an artificial neural network approach. ISIJ Int 56(2):183.
• He F, Zhang L (2018) Mold breakout prediction in slab continuous casting based on combined method of GA-BP neural network and logic rules. Int J Adv Manuf Technol 95:4081.
• Jin X, Li C, Wang Y, Li X, Xiang Y, Gu T (2020) Investigation and optimization of load distribution for tandem cold steel strip rolling process. Metals 10(5):677.
• Garcia-Mateo C, Capdevila C, Caballero FG, de Andrés CG (2007) Artificial neural network modeling for the prediction of critical transformation temperatures in steels. J Mater Sci 42:5391.
• Nürnberger F, Schaper M, Bach FW, Mozgova I, Kuznetsov K, Halikova A, Perederieieva O (2009) Prediction of continuous cooling diagrams for the precision forged tempering steel 50CrMo4 by means of artificial neural networks. Adv Mater Sci Eng 10:2009.
• Shah M, Das SK (2018) An artificial neural network model to predict the bainite plate thickness of nanostructured bainitic steels using an efficient network-learning algorithm. Adv Mater Sci Eng 27:5845.
• Lee SI, Shin SH, Hwang B (2021) Application of artificial neural network to the prediction of tensile properties in high-strength low-carbon bainitic steels. Metals 11(8):1314.
• Yemelyanov V, Yemelyanova N, Safonova M, Nedelkin A (2018) The neural network to determine the mechanical properties of the steels. In AIP Conf Proc 1952(1):020032.
• Saravanakumar P, Jothimani V, Sureshbabu L, Ayyappan S, Noorullah D, Venkatakrishnan PG (2012) Prediction of mechanical properties of low carbon steel in hot rolling process using artificial neural network model. Procedia Eng 38:3418.
• Somkuwar V (2013) Use of artificial neural network for predicting the mechanical property of low carbon steel. J Eng Comp App Sci 2(3):43.
• Fujita T, Ochi T, Tarui T (2007) Prediction of hardness distribution in forged steel by neural network model. Nippon Steel Tech Rep 96:57-61.
• Brownlee J (2020) Impact of dataset size on deep learning model skill and performance estimates. https://machinelearningmastery.com/impact-of-dataset-size-on-deep-learning-model-skill-and-performance-estimates/ Retrieved May 10, 2024.
• Rolinska M, Gustavsson F, Hedström P (2022) Revisiting the applications of the extraction replica sample preparation technique for analysis of precipitates in engineering alloys. Mater Charact 189:111978.
• Mukherjee T, Stumpf WE, Sellars CM (1968) Quantitative assessment of extraction replicas for particle analysis. J Mater Sci 3:127-135.
• Kisakurek SE (1986) On application of the carbon extraction replica technique for the determination of second phase particles dispersed into metal matrix. Metall 19(1):19-25.
• Jaiswal S (2024) Multilayer Perceptrons in Machine Learning: A Comprehensive Guide, https://www.datacamp.com/tutorial/multilayer-perceptrons-in-machine-learning. Retrieved May 10, 2024.
• Stathakis D (2009) How many hidden layers and nodes?. Int J Remote Sens 30(8):2133.
• Ding S, Li H, Su C, Yu J, Jin F (2013) Evolutionary artificial neural networks: a review. Artif Intell Rev 39(3):251.
• Kaastra I, Boyd M (1996) Designing a neural network for forecasting financial and economic time series. Neurocomputing 10(3):215-236.
• Hadzima-Nyarko M, Trinh SH (2022) Prediction of compressive strength of concrete at high heating conditions by using artificial neural network-based Bayesian regularization. J. Sci. Transp. Eng. 2(1):9-21.
• Zhang X, Sun L (2021) Optimization of optical machine structure by backpropagation neural network based on particle swarm optimization and Bayesian regularization algorithms. Materials 14(11):2998.
• Ackermann MA (2020) Bainitic TRIP Steels for Controlled Cooled Wire Rod, PhD Thesis. Universitätsbibliothek der RWTH, Aachen, Germany. 132 p.
• Keskin B (2019) Bainitic transformation in low carbon micro-alloyed hot forged steels for diesel engine components, MSc Thesis, Middle East Technical University, Ankara, Türkiye. 87 p.
• Zhou M, Xu G, Tian J, Hu H, Yuan Q (2017) Bainitic transformation and properties of low carbon carbide-free bainitic steels with Cr addition. Metals 7(7):263.
• Hudok D (1990) Properties and selection: irons, steels, and high-performance alloys, Metals handbook, ISBN: 978-0-87170-377-4.
• Ali M, Kaijalainen AJ, Hannula J, Porter DA, Kömi JI (2020) Influence of Chromium Content and Prior Deformation on the Continuous Cooling Transformation Diagram of Low-Carbon Bainitic Steels. Key Eng Mater 835:58-67.
• Yao Z, Xu G, Hu H, Yuan Q, Tian J, Zhou M (2019) Effect of Ni and Cr addition on transformation and properties of low-carbon bainitic steels. Trans Indian Inst Met 72:1167.
• Long X, Zhang F, Yang Z, Lv B. Study on microstructures and properties of carbide-free and carbide-bearing bainitic steels. Materials Science and Engineering: A. 2018 Feb 7;715:10-6.
• Kumar R, Dwivedi RK, Arya RK, Sonia P, Yadav AS, Saxena KK, Khan MI, Moussa SB (2023) Current development of carbide free bainitic and retained austenite on wear resistance in high silicon steel. J Mater Res Technol 24:9171-9202.
• Kaletin AY, Ryzhkov AG, Kaletina YV (2015) Enhancement of impact toughness of structural steels upon formation of carbide-free bainite. Phys Met Metallogr 116:109-114.
• Hu H, Xu G, Zhou M, Yuan Q (2016) Effect of Mo content on microstructure and property of low-carbon bainitic steels. Metals 6(8):173.
• Zhu M, Xu G, Zhou M, Hu H (2019) The Effects of Cooling Mode on the Properties of Ti–Nb Microalloyed High-strength Hot-rolled Steels. J Wuhan Univ Technol Mater Sci Ed 34(3):692.
• Wang H, Cao L, Li Y, Schneider M, Detemple E, Eggeler G (2021) Effect of cooling rate on the microstructure and mechanical properties of a low-carbon low-alloyed steel. J Mater Sci 56:11098.
• Altamirano G, Mejía I, Hernández-Expósito A, Cabrera JM (2012) Effect of boron on the continuous cooling transformation kinetics in a low carbon advanced ultra-high strength steel (A-UHSS). MRS Online Proc Lib 1485:83-88.
• Ali M, Nyo T, Kaijalainen A, Javaheri V, Tervo H, Hannula J, Somani M, Kömi J (2021) Incompatible effects of B and B+ Nb additions and inclusions' characteristics on the microstructures and mechanical properties of low-carbon steels. Mater Sci Eng A 819:141453.
• Da Rosa G, Maugis P, Portavoce A, Drillet J, Valle N, Lentzen E, Hoummada K (2020) Grain-boundary segregation of boron in high-strength steel studied by nano-SIMS and atom probe tomography. Acta Mater 182:226-234.
• Klein BD, Rossin D (1999) Data quality in linear regression models: Effect of errors in test data and errors in training data on predictive accuracy. Info Sci 2:33.
• Khamis A, Ismail Z, Haron K, Mohammed AT (2005) The effects of outliers data on neural network performance. J Appl Sci 5(8):1394.